Recognition memory is thought to be based on two processes, recollection and familiarity. The proposed research will investigate contrasting views of one of these two processes, namely, the recollection process. The question of how best to conceptualize recollection is of fundamental importance to computational models and neuropsychological models of memory. One view, which is by far the most common view, holds that recollection is a threshold process in that whenever it occurs, it carries with it high confidence that an item was seen before. This idea is explicit in some models, but it is also implicit in the thinking that underlies the popular Remember/Know procedure. An alternative view is that recollection is a continuous process (i.e., recollection comes in degrees, including very small degrees). This account is compatible with signal-detection theory, and it holds that recollection can be associated with low, medium or high degrees of confidence (and accuracy), depending on the degree of recollection associated with the test item. The resolution of high-profile debates in the cognitive neuroscience literature hinges to a large extent on a resolution of this question concerning the nature of the recollection process. Behavioral methods, including ROC analysis, relational memory procedures, and variations of the Remember/Know procedure will be used to address this problem. In several of these experiments, the crux of the issue is whether recollection occurs even when accuracy is low, confidence is low, and subjects claim that recollection absent. PUBLIC HEALTH RELEVANCE: Memory deficits are now recognized to be a core feature of several mental illnesses, especially schizophrenia. The methods that are typically used to further our understanding of these memory deficits are almost exclusively based on a model of memory that may not be valid. The proposed work tests that model of memory against an alternative and long-standing model known as signal-detection theory.